Urban Development

July 1, 2024

Urban Planning & AI: LLM Massing Generator

Sofia Malmsten

CEO & Architect

Urban planning has long been a complex field, requiring a delicate balance between architectural vision, regulatory compliance, and public interest. However, advancements in digital tools and data management are transforming this landscape. Notably, GPT-4 and large language models (LLMs) are opening new avenues for innovation, processes, and application areas. When AI can generate masses and we have access to the full design scope at our fingertips, design can now be used as a form of quality assurance in various processes. This article emphasizes new approaches to design as a consequence of AI development and access to massing generation tools.

What is a Massing Generator?

A massing generator is a digital tool used in urban planning and architecture to create preliminary volumetric representations of buildings and structures. These representations, known as massings, are essential for understanding the spatial impact of a proposed development within its context. Massing generators simplify the initial stages of design by providing quick and versatile models that can be adjusted and iterated upon rapidly. With the advancements in AI, particularly GPT-4, these tools can now be utilized in specialized applications for generating massing, akin to having a digital architecture colleague.

Hektar ai — massing generation tool for early stages of urban planning
Hektar — massing generation tool for early stages of urban planning

Massing Generation as enabler for New Processes

Having access to the design scope at your fingertips can impact several other processes, allowing us to utilize design in new ways. Instead of specifying exact buildings, planners can allow for flexible use for various functions and project types within the detailed plan. With AI, we can use massing generation tools to investigate this design scope. For example, tools as Hektar can benchmark and evaluate design scopes without the need to draw them all manually.

Flexible Detailed Plans

One process directly affected by this is the design of flexible detailed plans. Flexibility is a cornerstone of modern urban planning. With the ability to generate and modify detailed plans quickly, planners can respond to changing requirements and stakeholder feedback efficiently. Instead of seeing Hektar as just a design tool, planners can use Hektar as a colleague for feedback when setting requirements for zoning and detailed plans.

The ongoing trend towards more flexible plans presents a dilemma: the desire for flexibility while maintaining control. That’s where Hektar comes in. It allows planners to test constraints and see possible outcomes within the design requirements, ensuring that all possible designs within a flexible plan maintain high quality. Design is not just about one proposal; it can also be used for requirement specifications.

Example of how Hektar can be used for massing generation and evaluation of design KPIs
Example of how Hektar can be used for massing generation and evaluation of design KPIs

Where does AI comes into play?

To rely on the generated proposals, incorporating architectural knowledge is essential. Hektar, a SaaS platform for massing generation, is designed to be context-aware. Context awareness, requirements, and the combination of design strategies are complex tasks. How proposals interact with their surroundings, adhere to requirements, and optimize cost criteria and planning demands is very intricate. This is where LLMs, particularly GPT-4, come in. By combining GIS services and WMS information from geospatial systems, AI can support us in making decisions and seeing trends across the wide design scope. With tools like Hektar combined with AI, we can access and utilize this information in new and innovative ways.

Contextual aspects that influences the design
Contextual aspects that influences the design

Design principles relevant for AI models to consider in an urban configurator
Design principles relevant for AI models to consider in an urban configurator

Case Study: Eslöv Municipality

In an innovation project with Eslöv municipality, we tried the combination of generative design, GIS integration and AI. We generated several options for massing to explore what could be built within a given area and set of design constraints. Various typologies, such as point houses, quarters, and lamella structures, were combined with parking solutions to optimize the use of structures, surfaces, and typologies within the area. Over time, Hektar became better at recognizing the surroundings, such as the street noise on the eastern part of the plot and the park with apple trees to the right, incorporating more open structures accordingly.


The convergence of massing generators, access to open data, AI, and digital detailed plans is creating new processes and use cases in the planning industry. These tools offer unparalleled flexibility, detail, and efficiency, enabling planners to create more informed and dynamic urban spaces. We don’t even need to see the generated designs as the final projects to build but rather as benchmarks to validate our hypotheses and requirements for future cities.

As technology continues to evolve, the possibilities for smarter, more sustainable urban development will only expand, making our cities better places to live, work, and play. By embracing these modern tools, urban planners can navigate the complexities of city planning with greater ease and precision, ultimately contributing to the creation of vibrant and resilient urban environments.

Curious to know more? Read more and try our Hektar on your own projects here: